financial markets group - LSE

2 downloads 241 Views 327KB Size Report
(1998) argues that "once national authorities give up their command over money creation, they ... first group argues that fiscal rules are needed to discipline governments. ...... The Financial Planning Council is chaired by the Federal Minister of.

Managing Government Default Risk in Federal States By Jan Lemmen SPECIAL PAPER 116 August 1999

_______________________________________________

FINANCIAL MARKETS GROUP AN ESRC RESEARCH CENTRE _______________________________________________

LONDON SCHOOL OF ECONOMICS

Any opinions expressed are those of the author and not necessarily those of the Financial Markets Group. ISSN 1359-9151-116

Managing Government Default Risk in Federal States Jan Lemmen* Financial Markets Group London School of Economics

First version March 1999, Second version August 1999. Abstract

Federal governments typically apply fiscal rules to impose fiscal discipline on lower levels of government. Analogously, by trading in government debt, government bond markets impose fiscal discipline on lower levels of governments. This paper finds new evidence for Australia, Canada and Germany showing that whether these rules or markets matter, or not, may be a function of the world's appetite for credit risk. Rules and markets only tend to bite during periods when there is a low appetite for credit risk in world financial markets. Therefore, this paper proposes an alternative more incentivebased framework of fiscal discipline. This incentive-based framework should increase the sensitivity of government borrowing costs with respect to debt levels, increase the geographical diversification of investor's portfolios with respect to government bonds, and prevent government financing from fuelling private or public sector bailout expectations.

JEL Classification: H6, G18, G28.

Keywords: Government default risk, fiscal rules, government bond markets, federal states.

*

Mailing address: Financial Markets Group, London School of Economics, Houghton Street, London WC2A 2AE, United Kingdom, Tel: +44-171-9557051, Fax: +44-171-2421006, E-mail: [email protected], Web: http://cep.lse.ac.uk fmg/people/index.html. I thank Ivo Arnold, Graham Bishop, Charles Goodhart, Charlotte Ostergaard, Christian Pedersen and Niels Thygesen for valuable comments and discussion. The author is also grateful to Gerald Stuber and Micheline Roy (Bank of Canada), John Lester and Don Mikolich (CIBC Wood Gundy Securities), Peter Hall (Conference Board of Canada), Thomas Swartz (Chubb Corporation), and Richard Bennett and Peter Grossman (Queensland Treasury Corporation) for help with the data. I gratefully acknowledge financial support by the TMR Program of the European Commission (Contract no: ERBFMBICT972489). Any opinions expressed are those of the author.

1. Introduction Like state/provincial governments in federal states, national governments in EMU are expected to be constrained in the way they finance their deficit. The credit status of EMU governments will be analogous to the 50 states of the US or the 10 provinces of Canada. Government bonds issued by EMU member states will bear credit or default risk since no national government will have the sole authority for printing or creating Euros.1 Thus, national governments in EMU no longer can halt a process of falling bond prices and rising interest rates by monetising their debt. McKinnon (1997) argues that "when the national government owns its own central bank, everybody knows that, in a crisis, the government can always print money, that is use the inflation tax, to pay interest and principal and thus avoid outright default on the face value of its obligations. Because easy (potential) access to monetary seigniorage greatly reduces any risk of outright default, the government that owns the central bank can pre-empt the national capital market to issue Treasury securities at lower interest rates than can highquality private borrowers whose debt is denominated in the national currency." Similarly, Goodhart (1998) argues that "once national authorities give up their command over money creation, they lose the unchallenged absolute ability to pay off their domestic currency debt, interest and principal, in legal tender, whatever may happen to demand in the bond market." Finally, Bishop (1998, p. 17) argues: "With modern and efficient capital markets of a global scale, there will be no need for Euro investors to expose themselves unnecessarily to such a risk so prudent institutions may seek to diversify away from perceived risk. So a sudden liquidity crisis is entirely possible." This paper is about the management of default or credit risks in federal states. This involves measures that reduce the likelihood of government default. 2

Generally, there are two groups of thought on how to reduce the likelihood of government default. The first group argues that fiscal rules are needed to discipline governments. Fiscal rules such as deficit, debt or expenditures limits, a golden rule (that public borrowing must not exceed public investment) or a balanced budget law are theoretically justified if they help to avoid government debt accumulation, increasing the potential for a reduced default risk premium on government borrowing rates. I shall call

1

All domestic bonds issued by EMU member states will reflect the same currency and inflation risk. There have been no recent examples of actual state or provincial government default in the more advanced federal states. The most prominent cases of local governments experiencing fiscal problems were Orange County and New York City. On 6 December 1994, Orange County, one of California's most affluent regions, became the largest US municipality ever to file for bankruptcy; the move followed the revelation of investment losses of at least $ 1.5 billion and the refusal of several Wall Street banks to renew the municipality's short-term loans (Keesing's Record of World Events, p. 40315). The investment losses were due to speculation in the municipal bond market by the county treasurer. Speculation and the unwillingness to resume debt service rather than irresponsible fiscal behaviour caused the default. New York City came close to defaulting on some notes but eventually with the assistance of the State of New York it was able to continue debt service. 2

1

this the rule-based fiscal disciplining hypothesis. 3 The second group argues that markets discourage governments from running persistent deficits by demanding compensation for the additional probability of default that results from prolonged debt accumulation and eventually by limiting the availability of credit as borrowing becomes excessive (Bishop, Damrau and Miller, 1989). I shall call this the market-based fiscal disciplining hypothesis. 4 This paper will examine whether rule- and/or market-based fiscal discipline is effective in lowering government borrowing costs and whether alternatives are available.5 The paper makes the following contributions: US and other federal states Most of the existing literature that has tested for the rule- and market-based fiscal disciplining hypothesis draws on US states (see Table 1 in Section 2). It is not clear, however, whether the US evidence necessarily carries over to other federal states and to EMU. Indeed, the US may be a special case in part since municipal bonds are exempt from US federal income taxation. Because of this taxadvantage, the equilibrium nominal interest rates on states and municipal bonds are significantly less about 2 percentage points - than on US Treasury or high-grade corporate bonds (McKinnon, 1997). So, in the US resident individuals and corporations tend to hold more state debt in stead of banks and institutional investors. Foreigners, who would not receive the tax advantage, do not buy them, and the market becomes narrowly "onshore" in US dollars. This makes it easier for the US federal government to ignore bankruptcies in any one state or locality, i.e. no bailout, because such a bankruptcy would not impair the country's international credit rating, and, possibly the standing of the other states (McKinnon, 1997). However, in Europe (see Table 9 in Section 4) and other federal states such as Australia, Canada and Germany resident individuals and corporations do not receive this tax advantage. Banks and institutional investors often hold significant proportions of their own government's debt within their asset portfolios. The large credit exposures of banks and institutional investors to their own government could endanger the stability of the EMU financial system and thereby increase the pressure on the European Central Bank (ECB) or other governments to buy up the 3

A large literature in public finance has examined the effect of fiscal rules on deficits as opposed to the effect of fiscal rules on government borrowing costs. Von Hagen (1991, p. 209) concludes that fiscal rules have little effect on fiscal performance: "Fiscal restraints do not seem to significantly affect average fiscal performance other than the choice of debt instruments." But Von Hagen also runs non-parametric tests that reveal that formal fiscal restraints do have significant effects on the likelihood of per capita debt. Two other studies, ACIR (1987) and Poterba (1996) found that fiscal rules do matter in that they induce smaller deficits. In principle, EMU member states running deficits to GDP of more than 3% will be subject to sanctions (fines), though the Stability and Growth Pact (1996) establishes that EMU member states running excessive deficits of 3% or more will be automatically exempt from these sanctions in the event of a natural disaster or if they experience a fall in GDP of at least 2% over a year. In cases where GDP has fallen between 0.75% and 2% the imposition of sanctions will be dependent on political judgement (voting by EU finance ministers). Countries experiencing negative GDP growth of less than 0.75% will suffer sanctions if they run excessive budget deficits. 4 Bishop et al. (1989) argue that three conditions must be satisfied in order for market-based fiscal discipline to work: (1) capital must be able to move freely, (2) full information must be available on the creditworthiness, and the debts of the borrowers, and (3) the market must be convinced that there is no possibility of bail-out - that there are no formal or implicit guarantees that obligations will be met. 5 Note that there may be a trade-off between the objective of reducing government borrowing costs and the objective of containing risk, that is, insulating the government portfolio from interest rate shocks (Missale, 1998, p. 5).

2

troubled government's bonds. So it is important to test the robustness of the US evidence on new data sets. World's appetite for credit risk Whether rule and/or market discipline matters, or not, may be a function of the broader economic conjuncture as for example measured by the world's appetite for credit risk (see Appendix C). I argue that rules and markets bite during certain periods, but not during other periods. This follows from an inspection of the graphs of 10-year government bond yield differentials of state/provincial governments vis-à-vis the federal government and debt-to state GDP ratios in Appendix A. The development of yield differentials seems quite different from the development of debt-to-state-GDP ratios. This leads me to believe that there is something missing. I argue that the impact of rule- versus market-based fiscal discipline is connected with the world's appetite for credit risk. The development of state/provincial yield differentials with respect to the federal government is remarkably similar across federal states. State/provincial government bond yields in Australia, Canada, Germany, Switzerland 6 and the US 7 had dramatically dropped to the federal yield level just before the outbreak of the Asian crisis in 1997 and the Russian default of August 1998.8 In the aftermath of the Asian crisis and the Russian default differences between provincial and federal yields widened considerably. The Economist (1999) provides the following explanation: "After Russia's default, in particular, investors eschewed all kinds of risk. As in every financial crash since the 17th-century tulip crisis, greed suddenly turned to fear. The difference this time was that the effect was global, not just local. When markets panicked after Russia's crash, the impact was felt everywhere, from Brazil to American municipal bonds." I argue that this is due to shifts in investor's attitudes away from credit risk prompting a "global flight to quality" and thus increasing the yield differentials between lower levels of government and the federal government. Or in other words, provinces may find it more difficult to issue new bonds than the federal government due to a reduction in demand for "risky" provincial debt. During times of a low appetite for credit risk there is a perception that lower levels of government are more likely to default than federal governments. In the econometric tests the world's appetite for credit risk will be measured by the ratio of an index of emerging markets' sovereign debt (SALOMON BRADY BOND INDEX (USD)) and an index of G7 countries' sovereign debt (SALOMON G-7 INDEX (USD)) (in percentages). The paper finds new evidence for Australia, Canada and Germany showing that whether rules or markets matter, or not, may be a function of the world's appetite for

6

This follows from a plot of yield differentials of Swiss Cantons with respect to the Bund based on data obtained from Datastream (available upon request). 7 This follows from a plot of 20 Year General Obligation Bond Trading Values obtained from the Chubb Corporation (available upon request). 8 This is also partly due to budget tightening in Canada and the United States (McKinnon, 1997).

3

credit risk. Rules and markets only tend to bite during periods when there is a low appetite for credit risk in world financial markets. Political risks I shall also control for the most obvious state-specific political risks. For example, during state elections and referenda, large swings in yield differentials are observed. Specifically, the higher political risks associated with Québec are estimated by applying Gallup opinion poll survey data and election results showing the popularity of the Bloc Québécois. Budget rules The existing literature often applies a budget rules stringency index (for example, ranging from 1 to 10 or otherwise) to test for the rule-based fiscal disciplining hypothesis. To clear the budget stringency index of any spurious cardinality and to limit the extent of possible classification error, the use of dichotomous variables is advisable. It is better to transform the available data range into a dummy variable. By transforming every class in a zero-one variable no pre-estimate restrictions are put on the possible relationship between budget rules and yield differentials (Siermann, 1996). I include in my regressions a fiscal stringency dummy variable FISC (which takes the value of 1 for states during times when strong fiscal rules are in place and 0 otherwise). Liquidity risk A general drawback of the existing literature is that credit and liquidity risks are always lumped together. In addition to credit considerations, I expect liquidity to play a central role in the determination of yields in EMU government bond markets (see also McCauley, 1999 and Financial Times, 1999c).9 Liquidity risk is the risk that an investor may not be able to liquidate an investment in debt securities within a reasonable time at a reasonable price. Governments desire to minimise debt service costs. A liquid secondary market makes it easier for governments to issue large amounts of public debt at relatively low costs since investors feel more confident in their ability to purchase the issue in the primary market and subsequently trade the issue in a liquid secondary market. The smaller the size of the debt issue, the higher the liquidity premium a government has to pay to investors to buy its government bonds. I will assess differences in the level of liquidity across states/provinces by the relative amount of gross new domestic debt issues of each province/state over time. In this study, I am unable to assess how differences in the structure and organisation of state/provincial debt markets cause market liquidity to differ across states/provinces.

9

EMU government default risk premiums are very small (Lemmen and Goodhart, 1999).

4

The remainder of the paper is organised as follows. In Section 2 I begin with a review of the existing literature that has examined the rule- and market-based fiscal disciplining hypothesis. Following this review, Section 3 applies a simple model to examine the rule- and market-based fiscal disciplining hypothesis for Australia, Canada and Germany. The empirical testing of this takes the form of standard pooled and fixed-effects regressions.10 I control for the world's appetite for credit risk and the size of gross new domestic public debt issues over time. Section 3 includes also a brief description of each federal system and its credit implications. Section 4 argues in favour of a more incentive-based framework of fiscal discipline. I give some suggestions for imposing fiscal discipline on governments in the Euro zone. Finally, Section 5 concludes.

2. The literature on the rule- versus market-based fiscal disciplining hypothesis Table 1 summarises the literature and Appendix B reviews the theoretical model underlying the majority of the literature used to test the rule- versus market-based fiscal disciplining hypothesis. The market-based fiscal disciplining hypothesis suggests that financial markets first discipline with rising risk premiums reflecting the higher probability of default that results from an increase in the relative supply of debt (linear relationship between yield and debt). Second, if deficits persist, the default premium would increase at an increasing rate until, eventually, the offending country would be denied access to additional credit (non-linear relationship between yield and debt). The increase in the cost of borrowing, along with the threat of reduced availability of credit, would then provide an incentive to correct irresponsible fiscal behaviour.

Bishop, Damrau and Miller (1989) found that New York City borrowing expenses rose from 4% of the City's funds in 1965 to 53% in 1975. The spread between medium grade and prime long-term municipal bonds averaged between 40-50 basis points in 1974 and 60-70 basis points in 1975, and around 100 basis points at the height of the New York City's fiscal crisis. Though, within a year the spread had fallen back to 20 basis points.

Goldstein and Woglom (1991) and Bayoumi, Goldstein and Woglom (1991) also find empirical support for the market-based fiscal disciplining hypothesis. In addition, they find evidence for the rulebased fiscal disciplining hypothesis. Goldstein and Woglom (1991) examine a panel of 39 US states' general obligation bonds relative to New Jersey with comparable 20-year maturity and other characteristics obtained from the Chubb Corporation. Applying pooled OLS techniques, they estimate

10

Testing for individual states/provinces requires long time-series, which are only available for the United States.

5

a linear specification of default risk as in (8) (see Appendix B) over the sample period 1982-1990. They find that for US states the spread increases by 9.98 basis points for every percentage point increase in debt. They also find default risk differentials to decrease by 0.40 basis points with respect to the squared debt ratio as specified in (9). That is, the authors do not find evidence for credit rationing of US states. Finally, more stringent budget rules are found to decrease government borrowing costs by 2.01 basis points relative to states with average stringency.

Bayoumi, Goldstein and Woglom (1995) (henceforward BGW) generally confirm these results, using the non-linear specification in (7) for the supply curve of government debt. BGW find that for US states the spread increases by 18.91 basis points for every percentage point in additional debt to gross state product. Importantly, they find that borrowing costs rise at an increasing rate as the state debt-to-GDP ratio increases. Capital markets begin to ration credit to state governments at relatively low levels of debt to gross state product (8.7%) - low compared to the average EMU-11 debt-to-GDP ratios in 1998 (73.8%). They also find that the degree that the risk premiums change as stringency increases is dependent on the level of debt. The state debt-to-GDP level at which credit rationing begins is higher when the state has a more rigid fiscal constraint. At average levels of debt, varying stringency of budget constraints can lower risk premiums by over 4 basis points.

Alesina, De Broeck, Prati and Tabellini (1992) analyse, among others, the impact of an increase in debt-to-GDP ratios on default risk measured as the difference between government debt yields relative to corporate debt yields. They find that the government borrowing costs increase by 1.51 basis points for every percentage point in additional debt-to-GDP (in first differences) in a sample of 12 OECD countries during 1978-1989.

Poterba and Rueben (1997) find borrowing costs to increase by 92.1 basis points for every percentage point increase in debt-to-gross state product. In the same regression, lax anti-deficit rules increase borrowing costs by 8.42 basis points, a debt limit lowers borrowing costs by 5.38 basis points, binding expenditure limits lower borrowing costs by 7.08 basis points and binding revenue limits increase borrowing costs by 17.61 basis points.

Mattina and Delorne (1997) apply the linear specification (8) and the non-linear specification (9) to yield differentials of Ontario (1965-1996), Québec (1975-1996) and Nova Scotia (1965-1996) relative to British Columbia. The linear model indicates that the risk premium increases by 2.57 (Ontario), by 2.15 (Québec) and by 2.01 basis points (Nova Scotia) for every percentage point increase in provincial debt-to-GDP. The non-linear model indicates that for every percentage point increase in the squared 6

Study

Measure of default risk

government Impact of 1% change Impact of 1% change in Impact of 1% change in Impact of BGW debt- Impact of fiscal rule(s) debt to GSP/GDP squared debt to exponential debt to service GSP/GDP GSP/GDP (linear)

Bishop, Damrau New York City debt crisis and Miller of 1975, Spread between (1989) medium grade and prime long-term municipal bonds averaged Goldstein and 39 US states 20-year Woglom (1991) general obligation bond yields relative to New Jersey, 1982-1990 Alesina, De 12 OECD government Broeck, Prati debt yields relative to and Tabellini corporate debt yields (time (1992) to maturity not always equal), 1978-1989 Bayoumi, 38 US states tax exempt Goldstein and 20-year general obligation Woglom (1995) bond yields relative to New Jersey, 1981-1990

(non-linear)

(non-linear)

(non-linear)

(basis points, t-statistic in (basis points, t-statistic (basis points, t-statistic in (basis points, t-statistic in (basis points, t-statistic parentheses) in parentheses) parentheses) parentheses) in parentheses) Between 40-50 in 1974, 60-70 in 1975, and around 100 basis points at the height of the New York City crisis (mid1975) Pooled OLS: 9.98 (4.92) Pooled OLS: -0.40 Pooled OLS: -2.01 (1.55) (8.18) Reject non-linearity

ACIR stringency index ranging from 1 to 10

Pooled OLS: 1.51 (3.52) for first difference of debt-to- GDP

Pooled (1.97)

2SLS:

18.91

Pooled 2SLS: 0.11 (2.2)

Pooled (3.94)

2SLS:

-4.10

At mean debt/GSP levels, 1% increase in debt, increases yield by 23 bp.

Credit rationing at ACIR stringency index debt/GSP levels of 8.7%, ranging from 1 to 10 25% above the maximum debt/GSP level in the data (7.1%)

At one standard deviation above mean debt/GSP level, yield increases by 35 bp.

Accept non-linearity

Rueben (1997)

yields 20-year general obligation bond yields relative to New Jersey, 1973-1995

(1.29) (lax anti-deficit rules) Limit on issuing debt: -5.38 (-1.11) Binding expenditure limit -7.08 (-1.97) Binding revenue limit: 17.61 (-2.66)

Mattina and Ontario (1965-1996), OLS: 2.57 (0.04) Ontario Delorne (1997) Québec (1975-1996) and Nova Scotia (1965-1996) OLS: 2.15 (0.001) Nova bond yields relative to Scotia yields in British Columbia OLS: 2.01 (0.001) Québec Note: p-values between OLS: 0.10 2 parentheses. Debt Ontario

(0.028)

OLS: 0.04 (0.001) Debt 2 Nova Scotia OLS: 0.05 (0.0005) Debt 2 Québec

Le Flandreau, 15 European countries on Pooled IV: 5.645 (4.69) Cacheux and gold standard, spread of Zumer (1998) country=s government bond yield over the risk-free British consul, 1880-1913 Pooled IV: 1.732 (3.06)

Note: p-values between parentheses `

-0.398 (-1.36) Pooled IV: 1.449 (6.28)

-0.345 (-3.39) Pooled IV: 0.356 (10.18) -0.528 (-3.22) AOn [email protected] dummy variable taking the value of 1 if the

9

gold standard, and 0 otherwise Lemmen and 12 and 13 EU countries, Goodhart (1999) 10-year government debt yields less 10-year interest rate swap yields, subperiod 1992-1996

Fixed effects: 1.51 (5.81) for first difference of debt-to-GDP for 13 EU countries (excluding Greece and Luxemburg) Fixed effects: 1.14 (4.96) for first difference of debt-to-GDP for 12 EU countries (excluding Greece, Italy and Luxemburg)

Notes: OLS = Ordinary Least Squares Pooled IV = Pooled Instrumental Variables Pooled 2SLS = Pooled Two-Stage-Least-Squares bp. = basis points (1 basis point is 0.01%)

10

3. The explanation of the state-federal yield differential

Next, I analyse the impact of rule- and market-based fiscal discipline on state-federal yield differentials. The state-federal yield differential is measured by the difference between the yield on 10-year state/provincial government bonds and the yield on 10-year federal bonds. Thus, the default risk premium is defined in terms of the federal yield, which acts as the reference point. The default risk premium on federal government bonds should be low and stable in comparison to that of the states because of its favourable financial position, its ultimate control of money creation and its taxing autonomy.

The market-based fiscal disciplining hypothesis is modelled through DEBT (linear impact) and DEBT 2 (non-linear impact). The rule-based fiscal disciplining hypothesis is modelled through FISC , a dummy variable which takes the value of 1 for states/provinces during times when strong fiscal rules are in place and 0 otherwise. The vector X includes the world's appetite for credit risk ( APPETITE ) and the relative amount of gross new domestic debt issues for each province/state ( AIS ). APPETITE is measured by the ratio of an index of emerging markets' sovereign debt (SALOMON BRADY BOND INDEX (USD)) and an index of G7 countries' sovereign debt (SALOMON G-7 INDEX (USD)). If the indices move closer to each other (away from each other) the world's appetite for credit risk declines (increases). The narrower (wider) the difference between both indices, the lower (higher) the appetite for credit risk in world financial markets. AIS is measured as the amount of gross new domestic debt issues of each province/state in percentage of total of gross new domestic debt issues of all provinces/states.

For Australia and Germany, due to the unavailability of debt and/or GDP at a quarterly frequency, I necessarily have to rely on annual data. Since for Australia and Germany I also do not have information on fiscal rules I arrive at the following simple regression specification for Australia and Germany (see Appendix B for derivation): S i , t = β 0 + β1 DEBT i ,t + β 2 DEBT i2,t + β 3 APPETITEi ,t + β 4 AIS i , t −1 + ε i ,t where i indexes states/provinces and t indexes time.

For Canada the regression also includes a proxy for fiscal stringency (FISC ) and a proxy for the popularity of Bloc Québécois ( BQ ). BQ is measured by the ratio of the number of votes for Bloc Québécois in percentage of the total number of votes. The variable BQ relates to Québec only and is 11

zero for the other provinces. The availability of quarterly data for AIS allows me to use the sum of the gross new provincial debt issues in the previous four quarters in percentage of the total of all provincial issues to remove the simultaneity problem between debt and the amount of gross new debt issues (countries with higher debt ratios often have more liquid markets). I arrive at the following simple regression specification for Canada:

4

S i , t = β 0 + β 1 FISCi + β 2 DEBT i ,t + β 3 DEBT 2i ,t + β 4 APPETITEi , t + β5 BQ i , t + β 6 ∑ AIS i , t − j + ε i ,t j =1

The estimations of both specifications will be supplemented with a descriptive analysis of the particular features of federal systems that are believed to be relevant for the size of the default risk premium. The descriptive analysis serves two purposes; it increases our understanding of the working of existing federal systems, its credit implications and it highlights the differences with the federal system in EMU.

A final word of caution is in order here. There are a host of other variables that could contribute to the explanatory power of the default risk premium.11 The empirical model set out in this section necessary entails a balance between a parsimonious econometric specification and the risk of omitted variable bias. Debt in EMU may also not generate the same default risk premium for any given percentage of national debt as does state or provincial debt. There are important structural and institutional differences between federal states such as the US and Canada and EMU. Beside differences in taxation, labour mobility in EMU is much lower than in other federal states which makes it easier in the latter to discipline higher spending governments by moving to another jurisdictions, that is, the tax base in other federal states is more variable than in Europe. The federal government in the US (25% of aggregate US GDP) and other federal states is much larger than in Europe (1.3 % of aggregate EU GDP). The degree of fiscal federalism within the EMU is nowhere near as advanced as in other federal states. So, the effect of federal grants and aid programmes on default risk premiums is rather modest in the EU (Goodhart, 1997). The subsequent section reports on the estimation results.

11

For example, the dependency ratio - which expresses federal cash transfers as a proportion of total revenues measures state's vulnerability in meeting its financial obligations in the event of federal transfer restraints. The income-gearing ratio, which expresses debt-servicing charges as a proportion of total revenue, is another potential indicator of unsustainable borrowing. Unfortunately, these ratios tend to exhibit few innovations in the data-generating process, thereby limiting their usefulness as explanatory variables. Further, the relative small number of observations available from quarterly data (Canada) and annual data (Australia and Germany) necessarily requires a parsimonious econometric specification.

12

Australia The Australian federal system Australia is a federal state consisting of six states - New South Wales, Queensland, South-Australia, Tasmania, Victoria, Western-Australia - and two territories - the Northern Territory and the Australian Capital Territory. The financial relationship between the Commonwealth government and the states derives in part from the Australian states' limited ability to raise own-source revenues (OECD, 1997). While the states account for nearly 50% of total general government outlays, they are responsible for raising only 24% of taxation. The amount of own-source revenue raised by the Commonwealth is considerably larger than its own-purpose outlays (OECD, 1997). Fiscal patterns are characterised by large flows of funds from the federal level (which collects about three-quarters of all tax revenues) to the states. These revenue sharing transfers account for in excess of 50% of the states= total general government revenues.12

The Australian states' borrowings are regulated by the Australian Loan Council, an institution dominated by the Commonwealth. The aggregate borrowings of off-budget authorities are similarly regulated through limits on total new borrowings (OECD, 1997). The Loan Council de facto has control over the fiscal policies of each of the states. Global limits were imposed on all public sector borrowings by each state. International borrowing is further limited as percentage of each state=s global limit. In exchange for global limits on their borrowing, the Commonwealth government has granted the states' borrowing authorities access to international capital markets that had previously been closed to them. Credit implications: Australia The system of revenue sharing transfers provides the states with a secure, ongoing source of revenue that is largely independent of regional cyclical swings that may disproportionately affect one state. When combined with states' own, less income-sensitive tax bases, the revenues of the Australian states are not as closely linked to the states' economy as is often the case for the Canadian provinces and the US states. The comprehensive system of fiscal equalisation in Australia provides a safety net for any state experiencing a deterioration in its relative fiscal capacity. These arrangements provide particular benefits to the less populous, otherwise fiscally weaker states. The Australian federal system also provides scope for discretionary support from the Commonwealth. This support is not automatic, as the Commonwealth provides no explicit guarantee of state obligations. The absence of automatic Commonwealth support thus allows scope for credit 12

The Commonwealth raises revenue largely as a result of its monopoly of income tax collections.

13

differentiation among the states based on their relative economic and financial strengths. The Commonwealth nevertheless is a source of strong support for the states. The Commonwealth would be likely to work with a state experiencing fiscal difficulties to ensure ultimate repayments, for several reasons: (1) the Commonwealth historically has been the largest single creditor of the states and so retains a strong interest in preventing a default; (2) the existence of large Commonwealth funded Special Purpose Payments (SPP) programs administered by the states gives the Commonwealth a strong interest in maintaining the ongoing viability of state operations; and (3) the Commonwealth's own standing in financial markets remains linked with that of the states, with any failure or delay by a state in meeting debt servicing obligations likely to rebound adversely on Commonwealth's access to funds.

Table 2 shows the average yield differential of Australian states over the Commonwealth for bonds with a remaining maturity of close to 10 years. Due to this fact the data in Table 2 must be treated cautiously. Yield differentials have at times exceeded 100 basis points. Table 2: Australian states average yield differential with Commonwealth (in basis points)

1989 1990 1991 1992 1993 1994 1995 1996 January >97-June >97 1989 1990 1991 1992 1993 1994 1995 1996 January >97-June >97

New South Wales 10-year 89 56 68 42 30 34 30 24 25 Tasmania 116 87 129 103 38 26 32 26 11

Queensland 10-year 88 66 68 38 26 31 26 27 25 Victoria 99 74 105 90 60 44 39 24 17

South Australia 10-year 97 70 105 112 61 53a 21b .7c -20d Western Australia 102 83 101 63 28 34 31 20 10

Notes: a

Remaining maturity 9 years. Remaining maturity 8 years. c Remaining maturity 7 years. d Remaining maturity 6 years. b

Source: Own calculations from data obtained from Queensland Treasury Corporation for all Australian states. See also Appendix D.

14

Table 3 presents pooled and fixed-effects regressions of the non-linear model for Australia, based on annual data. New South Wales is taken as the reference-state in the fixed-effects regressions (3) and (4). Because of the limited number of observations on which the estimations are based the results have to be interpreted with care. For the Australian states I do not have access to the gross new issues of public debt. State and Territory debt security issuance since the mid 1980's has been concentrated in the so-called State Central Borrowing Authorities (CBA's). CBA's issue AUD denominated debt securities in Australia, and AUD and FX denominated debt securities offshore (e.g. euromarkets). Unfortunately these data are collected under the Census and Statistics Act which does not permit disclosure of data without consent of any of the states.

In regression (2) borrowing costs increase by 4.79 basis points for every percentage point increase in states debt-to-GDP and decreases by 0.06 basis points for every percentage point increase in squared state debt-to-GDP. The results in regression (2) of Table 3 indicate that a one percent higher appetite for credit risk in world financial markets lowers states borrowing costs by 0.72 basis points. The world's appetite for credit risk accounts for an important part of the cross-sectional variation in yield differentials. The results in all four regressions show that Australian states are not credit rationed. Table 3: The non-linear model for Australia: 1990-1996 Explanatory variables

(1)

Constant

(2)

(3)

(4)

*

-58.51 [36.24]

68.62 [36.90]

Queensland

-6.28 [11.62]

7.72 [9.58]

South Australia

-9.88 [12.10] 10.89 [23.39]

-1.16 [10.35] 12.53 [17.81]

16.62 [26.00]

18.50 [20.59]

4.79 [2.18]

7.55 [16.24] 3.28 [0.75]**

7.68 [12.48] 7.75 [1.20]**

-0.06 [0.03]**

-0.05 [0.02]**

-0.11 [0.02]**

Tasmania Victoria Western Australia **

DEBT i, t

**

6.45 [2.36]

**

DEBTi,2t

-0.08 [0.03]

-0.72 [0.11]**

APPETITE i, t

S.E. of R

Regression

2

N

-0.58 [0.13]**

28.92

23.46

30.43

25.03

0.11

0.42

0.02

0.33

42

42

42

42

Notes:

(1) Si ,t = β 0 + β 1 DEBTi , t + β 2 DEBT2i , t + ε i , t β

β

β

β

it

εi t

15

n

(3) S i , t =

∑α

i

DUM i + β1 DEBT i ,t + β 2 DEBT 2i , t + εi , t

i=1 n

(4) S i ,t =

∑ α DUM i

i

+ β1 DEBT i , t + β 2 DEBT 2i , t + β3 APPETITE i ,t + εi , t

i =1

* **

Significantly different from zero at 95 per cent level of confidence (one-sided test) Significantly different from zero at 99 per cent level of confidence (one-sided test)

White heteroskedasticity-consistent standard errors are between brackets. Source: See Appendix D.

Canada The Canadian federal system Canada is a federal state of ten provinces (Alberta, British Columbia, Manitoba, New Brunswick, Newfoundland, Nova Scotia, Ontario, Prince Edward Island, Québec, Saskatchewan) and two territories (the Yukon Territory and the Northwest Territory). As a general principle, the provincial and federal levels have equal powers in raising taxes. All provinces have the right to impose their own income tax. In general, they top-up federal taxes on a percentage basis. Provinces also have the right to raise taxes through social insurance charges, and through indirect taxation (sales taxes). Three of the provinces have their own collection system for corporate taxes. All other provinces raise corporate income taxes using the federal corporate income tax structure and administration (OECD, 1997). The Federal government has since the start of the Confederation made transfer payments to provinces to ensure equity in public services across the provinces and to allow them to carry out their responsibilities. Federal transfers to provinces and territories constitute an important source of provincial revenue and have expanded enormously over the last decades (OECD, 1997). Credit implications: Canada In Canada provincial governments have more power to act independently with respect to expenditures and taxes. Both the Canadian and US federations are more market oriented. Credit ratings exert considerable market discipline on the provinces and the states. Interestingly, Canada unites market discipline with a successful and wide-ranging system of monetary transfers to the provinces - without being seen as an implicit guarantee of provinces' budgetary deficits (Bishop, Damrau and Millar, 1989).

The provincial governments in Alberta, Manitoba, New Brunswick, Nova Scotia, Québec and Saskatchewan have recently introduced restrictions on fiscal activities to control debt accumulation. 16

I am interested in the effect of these budget constraints on provincial borrowing costs. It is unclear if borrowing costs have decreased for the provinces with legislated budget constraints compared to those provinces without legislated budget constraints (note that most of the legislation was passed in 1995 and 1996). In 1997, the average yield differential over 10-year government of Canada benchmark bonds for the provinces with legislated budget constraints (Alberta (8), Manitoba (13), New Brunswick (16), Nova Scotia (23), Québec (39) and Saskatchewan (16), which is an average of 19) was only moderately lower (7 basis points) than the average yield differential for the provinces without legislated budget constraints (Newfoundland (33), Ontario (17) and Prince Edward Island (27), which is an average of 26). Over the first eleven months of 1998, the average yield for the provinces with legislated budget constraints (Alberta (17), Manitoba (25), New Brunswick (26), Nova Scotia (34), Québec (47) and Saskatchewan (25), which is an average of 29) was again only 7 basis points lower than the average yield differential for the provinces without legislated budget constraints (Newfoundland (43), Ontario (27) and Prince Edward Island (39) which is an average of 36). The above calculations are based on Table 4. Also for Canada, yield differentials have at times exceeded 100 basis points. Table 4: Canadian provinces average yield differential with federal government (bid side) (in basis points)

6/92-12/92 1993 1994 1995 1996 1997 1/98-11/98

2-year 26 13 15 11 7 4 7

6/92-12/92 1993 1994 1995 1996 1997 1/98-11/98

40 22 18 17 10 7 11

6/92-12/92 1993 1994 1995 1996

79 59 38 50 35

Alberta 3-year 5-year 10-year 28 40 55 25 35 50 18 19 28 13 16 18 8 9 11 4 6 8 8 11 17 Manitoba 43 59 74 35 51 65 21 29 40 20 23 29 9 10 18 7 10 13 14 17 25 Newfoundland 82 101 119 78 101 119 41 61 83 52 57 75 34 36 52

30-year 47 45 36 24 19 12 23

2-year 29 15 15 11 8 7 12

68 59 48 37 28 21 35

41 22 19 18 12 7 11

119 116 95 84 62

51 34 29 31 21

British Columbia 3-year 5-year 10-year 30-year 30 44 57 50 26 36 51 46 17 20 29 36 13 16 20 25 9 10 15 23 7 9 15 21 15 19 29 40 New Brunswick 44 60 74 68 34 50 64 59 22 30 43 50 20 24 33 40 11 13 23 33 8 11 16 22 14 18 26 35 Nova Scotia 54 71 88 84 50 69 85 83 32 44 64 75 33 37 49 58 20 21 32 42

17

6/92-12/92 1993 1994 1995 1996 1997 1/98-11/98

69 49 34 37 24 14 17

8/88-12/88 1989 1990 1991 6/92-12/92

39

1993 1994 1995 1996 1997 1/98-11/98

26 20 19 12 7 11

Prince Edward Island 72 88 104 70 87 106 37 52 73 40 44 58 23 24 39 15 20 27 22 28 39 Ontario 39 45 39 45 50 55 60 58 61 73 42 (1992) (1992) 41 58 73 24 34 49 22 26 35 11 12 23 8 11 17 15 18 27

101 101 81 67 49 37 54

67 51 28 23 13 7 11

70 71 32 26 12 7 14

66

50

55

68 58 43 33 25 39

29 29 28 17 14 17

45 33 34 19 15 20

Saskatchewan 87 109 88 108 45 64 30 39 13 24 11 16 17 25 Québec 73 93 71 85 79 97 78 89 75 91 (1992) (1992) 62 87 49 83 52 72 30 47 24 39 28 47

106 103 72 48 34 23 35

92 85 90 80 59 46 60

Source: Own calculations with weekly data obtained from CIBC Wood Gundy Securities. See also Appendix D.

Table 5 summarises the characteristics of Canadian provincial budget rules. The budget stringency dummy variable FISC takes the value of 1 for states during times when strong fiscal rules are in place (Alberta, Manitoba and Québec), and 0 when weak or no fiscal rules are in place. A more formal analysis of provincial-federal yield differentials is needed to determine whether budget constraints legislated by the provinces have lowered provincial government borrowing costs. Table 5: The characteristics of Canadian provincial budget rules Province

Alberta British Columbia Manitoba New Brunswick Newfoundland Nova Scotia Ontario Prince Edward Island Québec Saskatchewan

Budget Stringency Dummy variable (FISC) 1 0

In effect since:

Apply to realised deficits

Concrete debt elimination provision

Singleyear budget period

Penalties for not achieving balance

Referendum Escape requirement clauses for tax changes

1993 -

Yes -

Yes -

Yes -

No -

Nod -

Yes -

1 0 0 0 0 0

1995 1993 1996 -

Yes Yes Noa -

Yes Nob No -

Yes Noc Yes -

Yes No No -

Yes No No -

Yes Yes Yes -

1 0

1996 1995

Yes No

No No

Yes Nob

No No

No No

Yes Yes

18

Notes: a

Realised expenditures may not exceed budgeted expenditures by more than one per cent. Realised deficits must be offset in following fiscal year. b New Brunswick plans to introduce debt-elimination provisions in the near future. c Four-year budget period. d Alberta's provincial government has announced its intention to introduce such a requirement. The province currently has a referendum requirement for the introduction of a retail sales tax. e The premier of Ontario recently stated his government's intention to introduce balanced-budget legislation in the future that would also require referendum approval for any legislated tax increases. In the meantime, the province intends to cut the annual salary of its cabinet ministers by 25 per cent if the yearly budget targets announced in the 1996 budget are not achieved. Source: Millar (1997).

Table 6 presents pooled and fixed-effects regressions of the non-linear model for Canada. The vector of variables X in the first regression includes a fiscal dummy variable FISC . I control for higher political risks associated with Québec ( BQ ), the world's appetite for credit risk ( APP ), and 4

the relative amount of gross new provincial debt issues ( ∑ AIS i , t − j ). In the fixed-effects regressions j =1

(3) and (4) Ontario is taken as the reference-state.

Regressions (1) to (4) do not indicate that Canadian provinces are credit rationed. This contrasts with findings of Mattina and Delorne (1997). Provincial borrowing costs increase with 0.75, 0.62, 1.22 and 1.95 basis points for every percentage point increase in debt to provincial GDP decreases by 0.006. 0.004, 0.01 and 0.01 basis points for every percentage point increase in squared provincial debt-to-GDP (regressions (1) to (4)). Fiscal rules matter, but less so when the world's appetite for credit risk is included in the regression (compare regression (1) with regression (2) and regression (3) with regression (4)). An increase in the world's appetite for credit risk lowers borrowing costs with 0.41 (regression (2)) or 0.32 basis points (regression (4)). The political risks associated with Québec notably increase provincial borrowing costs by 3.68 (regression (2)) basis points for every one percent increase in the Bloc Québécois' share of the aggregate Canadian vote. There is clearly more risk relative to Ontario associated with Québec and Manitoba. I also ran regressions excluding Québec from the sample. The coefficient estimates for regressions (2)-(4) were of the same magnitude as those in Table 6. Only for regression (1) the coefficient for debt and debt squared were found to be different: 0.02 [0.38] and -0.0001 [0.003]. Finally, I obtain a positive coefficient estimate for the sum of the gross new debt issues in the previous four quarters in percentage of the total of all provincial issues. This result indicates that an increase in the liquidity of the provincial government bond markets increases provincial borrowing costs, which contrasts with a priori expectations. Apparently, the sum of the gross new debt issues in the previous four quarters in percentage of the total of all provincial issues may not be a very good way of measuring liquidity.

19

Table 6: The non-linear model for Canada: 1992Q3-1997Q4 Explanatory variables

(3)

(4)

Alberta British Columbia

30.32 [5.19]**

42.98 [5.88]**

-3.16 [6.35]

15.34 [7.44]**

Manitoba New Brunswick

28.42 [16.68]* -9.76 [14.01]

60.86 [16.11]** 23.63 [15.00]

Newfoundland

-11.01 [13.50]

23.28 [15.51]

1.25 [12.04] -5.78 [5.79]

30.21 [15.10]** 25.24 [9.10]**

Quebec

39.72 [6.27]**

64.02 [13.64]**

Saskatchewan

13.80 [12.42]

40.45 [14.91]**

Constant

(1)

(2) **

20.82 [10.25]

**

77.85 [8.38]

Nova Scotia Prince Edward Island

-14.13 [2.90]**

-8.50 [2.50]**

-35.39 [2.74]**

-22.95 [2.80]**

DEBT i ,t

0.75 [0.35]**

0.62 [0.33]*

1.22 [0.17]**

1.95 [0.32]**

DEBTi,2t

-0.006 [0.002]**

-0.004 [0.002]**

-0.01 [0.002]**

-0.01 [0.002]**

FISC i

APPETITE i,t

-0.41 [0.04]**

-0.32 [0.04]**

20

Germany The German Federal System Germany is a federal state that consists of 16 Länder (states and city states): Berlin, Brandenburg, Bremen, Hamburg, Hessen, Mecklenburg-Vorpommern, Niedersachsen, Nordrhein-Westfalen, Rheinland-Phalz, Saarland, Sachsen, Sachsen-Anhalt, Schleswig-Holstein and Thüringen. The federal system of Germany represents a strong commitment to federation, primarily because of provisions in the Basic Law of 1969 (OECD, 1997). The Basic Law exerts significant control over the budgetary policies of the Länder, through a Financial Planning Council that attempts to coordinate overall fiscal policy. The Financial Planning Council is chaired by the Federal Minister of Finance and consists of the Länder Ministers of Finance, the Minister of Economics, and representatives of the municipalities and the Bundesbank. The Financial Planning Council makes recommendations on the co-ordination of budgets and financial plans of the Bund (federation) and the Länder, although it cannot make any binding decisions. A major factor in negotiations between the Bund and the Länder is that the negotiating partners are unequal. On one side is the Bund and on the other the 16 Länder, a majority of which must agree on a solution. It is therefor usually easier for the Länder to get their way. The Basic Law, provides also that there must be

reasonable

equalisation between the financial weaker and stronger Länder, taking into consideration the financial capacities and financial requirements of the municipalities or associations of municipalities.13 This is ensured by the distribution of tax revenues and by a system of supplementary allocation.

Some 72 per cent of Land revenues are made up of Land taxes, such as capital tax, inheritance tax, and so-called joint taxes (whose revenue are shared by the Federation and the Länder, and sometimes also by the municipalities), such as turnover tax, corporation tax and income tax. The Land taxes account for approximately 13.5 per cent of Land revenue, income and corporation tax for approximately 53 per cent and turnover tax for some 27 per cent. Together they account for about 75% of all tax revenues (OECD, 1997). Transfers account for another considerable share of Land revenues. Within the framework of co-operative federalism, the Bund funds investments made by the Länder or the local authorities. But by doing so, the Bund can exert considerable influence on Länder and municipality policy. In the case of joint tasks, the Bund and the Länder share the financial burden equally.

13

As from 1995 the eastern Länder have been fully tied into the federal equalisation system.

21

Credit implications: Germany The Länder can borrow, and the Federal Government is not liable for their debts. However, the unique structure of the federal system provides the Länder with a level of credit safety very close to that of the federal government. The fiscal relationship between the federal government and the Länder is so tight that markets forces scarcely distinguish between the Länder. Differences in the Länder" borrowing terms in the domestic market are determined more by liquidity than by creditworthiness considerations.

Table 7 presents average yield differentials of the German Länder with the Bund (in basis points). The average yield differentials are very small. Liquidity risk considerations in stead of credit risk considerations are the main explanation for yield differentials. Länder governments have few bonds outstanding and seldom issue new bonds. Table 7: German Länder average yield differential with Bund (in basis points) BadenWürttemberg

Bayern

Berlin

Brandenburg

1993

10-year 33

10-year 20

10-year 24

10-year 21

1994 1995

27 22

26 23

24 24

26 25

1996 1997

12 14

16 8

22 12

22 14

1993

Bremen 19

Hamburg 17

Hessen 21

MecklenburgVorpommern 23

1994 1995

27 25

19 23

34 29

29 23

1996 1997

20 13

23 16

32 22

24 20

Niedersachsen

NordrheinWestfalen

RheinlandPhalz

Saarland

1993 1994

8 25

18 33

15 17

25

1995

22

27

23

24

1996 1997

14 6

18 15

15 18

21 12

SachsenAnhalt

SchleswigHolstein

Thüringen

1993 1994

Sachsen 22 16

34 25

23 13

24 48

1995 1996

19 21

26 28

21 22

30 22

1997

21

14

21

7

Source: Own calculations with monthly data obtained from Datastream. See also Appendix D.

22

Table 8 presents pooled and fixed-effects regressions of the non-linear model for Germany, based on annual data. Nordrhein-Westfalen is taken as the reference-state in regressions (3) and (4). The estimations have to be interpreted with care because of the limited number of data (annual frequency). The results in regression (2) of Table 8 indicate that one unit increase in the appetite for credit risk in world financial markets lowers the Länder borrowing costs by 0.15 basis points. Borrowing costs increase by 0.89 basis points for every percentage point increase in Länder debt-toGDP and decreases by 0.015 basis points for every percentage point increase in squared debt-toGDP (regression (2)). This means that there is no credit rationing taking place in all regressions. Notably, a percentage point increase in the relative amount of gross new debt issues lowers borrowing costs of the Länder with 0.19 basis points in regression (2) and 0.42 basis points in regression (4). Table 8: The non-linear model for Germany: 1994-1996 Explanatory variables

(3)

(4)

-2.05 [5.86]

12.69 [2.13]**

Bayern

2.78 [4.96]

21.06 [4.74]**

Berlin

-0.05 [4.30]

-2.77 [2.53]

Brandenburg

0.003 [3.93]

Constant Baden- Wuerttemberg

(1)

(2) **

16.17 [5.91]

**

37.04 [7.06]

Bremen

-2.48 [1.78] *

28.25 [16.31]

3.43 [9.16]

Hamburg

-3.78 [4.63]

4.60 [1.68] **

Hessen

7.64 [3.42]** 0.32 [3.13]

17 22 [3.36]** 1.23 [0.82]

Mecklenburg - Vorpommern Niedersachsen Rheinland - Phalz Saarland Sachsen Sachsen - Anhalt Schlewig - Holstein

Thueringen DEBT i, t

0.56 [0.45]

DEBTi,2t

-0.001 [0.008]

**

0.89 [0.37]

-0.015 [0.006]**

-5.62 [5.15]

10.70 [12.38]

-6.37 [4.42] 13.57 [10.78]

-2.59 [2.53]** -0.59 [2.49]

-5.61 [4.11]

7.35 [3.78]*

2.95 [4.42] 7.39 [4.06]*

-3.59 [5.61] 7.22 [6.09]

6.63 [8.60]

-4.08 [5.31]

**

2.42 [0.45]

3.65 [0.40]**

-0.06 [0.02]**

-0.06 [0.01]**

APPETITE i,t

-0.15 [0.04]**

-0.18 [0.04]**

AIS

-0.19 [0.06]**

-0.42 [0.26]

i ,t − 1

S.E. of

R

2

N

Regression

6.44

4.38

5.86

3.68

-0.01

0.43

0.16

0.60

48

32

48

32

Notes:

(1) Si , t = β 0 + β 1 DEBTi , t + β 2 DEBT2i ,t + ε i ,t

23

(2) Si ,t = β 0 + β 1 DEBTi , t + β 2 DEBT2i ,t + β 3 APPETITEi ,t + β 4 AISi , t −1 + ε i ,t n

(3) S i ,t =

∑α

i

DUM i + β1 DEBT i ,t + β2 DEBT 2i , t + εi , t

i=1 n

(4) S i ,t =

∑ α DUM i

i

+ β1 DEBT i , t + β 2 DEBT 2i ,t + β3 APPETITE i ,t + β4 AIS i , t −1 + εi , t

i =1

* **

Significantly different from zero at 95 per cent level of confidence (one-sided test) Significantly different from zero at 99 per cent level of confidence (one-sided test)

White heteroskedasticity-consistent standard errors are between brackets. Source: See Appendix D.

4. An incentive-based framework of fiscal discipline

The results in the previous section indicate that rules and markets only tend to bite during periods when there is a low appetite for credit risk in world financial markets. Therefore, I propose an alternative more incentive-based framework of fiscal discipline. Measures that act as an incentive for governments to improve their fiscal behaviour include:14 (1) The development of integrated government bond markets At present national provisions regulate the public offering, marketing and distribution of government debt.16 Banks that provide investment services must be subject to authorisation by their home country. This discourages issuers and their intermediaries from placing bonds on a crossborder basis, and therefore prevents investors from diversifying their portfolios. The Euro will highlight the existence of these regulatory barriers. The underlying dynamics of the liquidity and default risk situation in government bond markets will be essential for the geographical diversification of government bonds. Lowering transaction and administration costs, harmonising trading rules, market conventions and selling techniques (auctions, syndication) will increase the potential for diversification of government securities. Cross-border acceptability of prospectuses, listings, financial reporting, rules applying to investment service providers could help to decrease the home bias in government bond markets (see also the CEPS report by Lannoo and Gros, 1998). 14

Part of these incentive measures are adapted from reports by the World Bank on Brazil (1995) and Argentina (1996a,b), the IMF=s Code on Good Practices on Fiscal Transparency - Declaration on Principles (16 April 1998) and the IMF=s Draft Manual on Fiscal Transparency (19 October 1998) and a report by the CEPS Working Party on Capital Markets and EMU by Karel Lannoo and Daniel Gros. 15 There is the tendency of investors to overweight their portfolio with greater than expected allocations of domestic financial assets. Foreign assets are under weighted because of (1) the higher barriers and costs of foreign investment such as transaction costs, taxes, capital controls, market access, time zone differences, language barriers, administrative costs and so forth and (2) the higher risks of foreign investments such as political risks. 16 The international federation of securities regulators (IOSCO) has agreed on a set of standards for cross-border offerings of securities.

24

Inviting more foreign participation increases the cross-border placing of government bonds and would promote the diversification of government bond portfolios. With widely held bond portfolios the systemic risks of a fiscal crisis would be less severe, and the bailout pressures on the ECB or other EMU governments less strong (Arnold and Lemmen, 1999).

Recently, more attention has been focused on accounting and disclosure diversity to determine whether it is a significant barrier to international investment and whether public policy reforms are advisable (European Commission, 1998a). The International Accounting Standards (IAS) Committee has developed a set of global accounting standards. A common accounting system for all levels of government of EMU member states, with independent external auditing assigned to a private firm of accountants whose results are made public should be set up. The IMF has also set up a "Special Data Dissemination Standard" which provides standards for participating countries in disseminating economic and financial statistics. (2) Regulating borrowing from the private sector Commercial banks have often lend more to (lower levels of) governments than would be justified by the governments' own financial condition, in the belief that the debt carried an implicit or explicit government guarantee. Banks often hold government bonds at below-market interest rates, undermining the profitability of banks (Arnold, 1999).

The ECB (1999) calculates (see Table 9) that the relative importance of banking sector=s direct lending to governments in terms of total domestic assets decreased over the period 1995 to 1997 in most EU countries due to the Stability and Growth Pact (1996). Arnold and Lemmen (1999) however argue that the risks of existing large exposures of individual banks to public debt in several European countries remain high. The authors examine the sensitivity of a banks' position in public debt to government default shock and conclude that regulators should promote the geographical diversification of public debt held by banks since it would reduce the risks of bank failures following a fiscal crisis in the EMU. Regulators should apply the large exposure rule and the capital adequacy directive to public debt (Bishop, Damrau and Miller, 1989, Arnold and Lemmen, 1999). At present, the large exposure rule that states that a bank cannot lend more than 25 per cent of its capital to a single borrower does not apply to government debt. Banks and other financial institutions should disclose their credit exposure to all levels of government.

25

Table 9: Banks' domestic claims on the government sector (as percentage of banks' total domestic assets) Belgium Greece Spain Italy Germany The Netherlands Portugal France Denmark Sweden Finland Ireland United Kingdom Luxemburg

1992 26.12 24.34 14.61 12.98 12.42 11.4 19.73 2.74 12.25 6.59 3.21 n.a. 1.47 n.a.

1995 26.18 25.54 17.75 16.91 14.61 10.96 15.1 5.5 11.4 14.52 10.48 7.89 2.12 0.21

1996 24.5 22.1 18.28 17.38 13.85 10.17 12.72 6.76 8.62 6.64 n.a. 6.39 1.81 0.2

1997 22.85 n.a. 15.91 15.83 13.27 8.88 8.39 6.78 6.45 4.86 n.a. 2.77 1.28

n.a.

Note: n.a. non available. Source: ECB (1999, Table 2.13a).

In addition, the zero risk-weighting of bank lending to OECD governments in the current Basle capital adequacy standards is no longer appropriate.17 This would improve the incentives of the banking system to avoid risky loans to the more exposed governments. The problem with this proposal is that non-zero risk weighting would increase banks' capital needs, possibly lowering the banks' demand for public debt and depress government bond prices. (3) Prohibit governments from being shareholder of commercial banks and companies Governments often acquire equity in companies or commercial banks. Regulation should prohibit governments from having their own banks or companies. Such "government-owned" banks and companies have access to cheaper funding than comparable privately owned banks and companies because of the implicit or explicit government backing. Government-owned banks are encouraged to undertake more risky projects and tend to lend out a larger share of their assets to local, provincial and central governments than commercial banks (Arnold, 1999). (4) Harmonisation of differences in tax Differences in tax influence investor’s demand for public debt. Governments may want to establish tax incentives to the nonfinancial sector (households and corporations) both located at home (resident interest withholding taxes) and abroad (non-resident interest withholding taxes) to lower the exposure to the domestic financial sector, but this would presumably run foul of desires for EU tax harmonisation. Indeed, a major problem is whether, and how long, existing tax breaks for 26

nationals investing in own-country government bonds can continue, and what would be the fall-out if such discriminating tax advantages were required to be (progressively) removed. These tax breaks are probably not large enough to account fully for the extent of the home country bias in government bond markets.

(5) Internal government reform The above four incentive measures are not a sufficient condition for the prevention of fiscal crises. Investors face the risk that a government may not have the revenue and expenditure flexibility to respond to changes in the external environment, or that the government will choose not to, even if it has the ability to pay debt service. Every new administration starts with a new agenda, which often includes reneging on the commitments of its predecessors. The uncertain path for taxes and expenditures affects the government borrowing needs and thus its exposure to variations in interest rates. Internal control systems in governments should be adapted to increase governments= ability and willingness to pay debt service. Governments should implement modern risk-management techniques. Government should conduct projections of expenditures including a breakdown between capital and current spending and revenues, a breakdown of types of revenues and whether they are shared with other levels of government. In addition, predictions of the level and variability of cash flows to pay off debt service should be made. EMU governments should reduce their debt to levels that would be payable through the entire range of interest rate changes, government default shocks and output reductions that can be expected.

For example, take the case of Brazil. Government default would have severe consequences for the Brazilian banking system. The Brazilian government owes R$320 billion domestic debt to Brazilian banks and other financial institutions. Holdings of government paper amount to between 20 and 30 percent of the banking system=s assets. The recent currency depreciation vis-à-vis the US dollar has increased the stock of the debt with R$60 billion. About half of the debt falls due this year. Already there are doubts about whether investors will agree to roll it over. Some government creditors could ultimately conclude that the interest rates - no matter how high - would not compensate for the risk of holding government paper. Even a 10% fall in the value of the paper would be enough to wipe out the entire financial sector=s profits of last year (Financial Times, February 5, 1999, p. 16).

5. Conclusion

This paper relates the experience of three federal states (Australia, Canada, Germany) with rule- and market-based fiscal discipline to issues involving EMU. Given the nearly exclusive reliance of most 27

studies on US data, I wonder if the US evidence necessarily holds true for other federal states. The paper analysed by how much (in terms of basis points) state/provincial borrowing costs increase for every percentage point in additional debt or debt squared (market-based fiscal discipline) and/or by how much state/provincial borrowing costs decrease with more stringent budget rules (rule-based fiscal discipline).

Yield differentials for Australian states and Canadian provinces have at times exceeded 100 basis points, but are found to be small for the German Länder. These larger yield differentials are probably due to the relatively larger cross-sectional variation of credit risk premiums in Australia and Canada than in Germany. The measure for liquidity (the relative amount of gross new domestic debt issues) did not always point in the same direction. I could not find a systematic negative relation between issue size and yield differentials for the Canadian provinces. My measure for liquidity - the sum of the gross new debt issues in the previous four quarters in percentage of the total of all provincial issues - may not be a very good way of measuring liquidity. Fortunately, for Germany I am able to provide evidence that a higher liquidity reduces borrowing costs of the Länder.

My main finding, however, is that the world's appetite for credit risk is a key determinant of provincial-federal yield differentials. High provincial-federal yield differentials reflect a low world appetite for credit risk. Rules and markets only tend to bite during periods when there is a low appetite for credit risk in world financial markets. Particularly since the Russian default of August 1998, the world's appetite for credit risk is the driving force behind yield differentials of provincial governments vis-à-vis the federal government. Future developments in Brazil or Japan are more likely to influence provincial-federal yield differentials than domestic factors such as provincial debt-to-GDP ratios or fiscal rules. Even so, this does not mean that state-specific (political) can be safely disregarded. The case of Québec shows that state-specific political risks may independently explain an important part of provincial-federal yield differentials.

Finally, this paper proposes an alternative more incentive-based framework of fiscal discipline. This incentive-based framework should increase the sensitivity of government borrowing costs with respect to debt levels, increase the geographical diversification of investor's portfolios with respect to government bonds, and prevent government financing from fuelling private or public sector bailout expectations. The overall result is likely to be a strengthening of market discipline and a reduction in the aggregate level of public debt in the financial system.

28

References ACIR (1987), Fiscal Discipline in the Federal System: National Reform and the Experience of the States, Washington. Arnold, I.J.M. (1999), AThe Third Leg of the Stool, Financial Stability as a Prerequisite for [email protected], Weltwirtschaftliches Archiv, Vol. 135, No. 2, Forthcoming. Arnold, I.J.M and J.J.G. Lemmen (1999), “The Vulnerability of Banks to Government Default Risk in the EMU”, FMG Special Paper, London School of Economics, London, Forthcoming. Battley, N. (ed.) (1997), The European Bond Markets: An Overview and Analysis for Issuers and Investors, Sixth Edition, The McGraw-Hill Company, London. Bayoumi, T., M. Goldstein and G. Woglom (1995), ADo Credit Markets Discipline Sovereign Borrowers?, Journal of Money, Credit and Banking, Vol. 27, No. 4, pp. 1046-1059. Bishop, G. (1990), A1992 and Beyond, Creating an EC Monetary Union with Binding Market [email protected], Salomon Brothers, February. Bishop, G. (1998), AThe European Central Bank and the Prudential Regulation of the Financial [email protected], International Journal of Finance and Economics, Vol. 3, pp. 13-20. Bishop, G., D. Damrau and M. Miller (1989), AMarket Discipline CAN Work in the EC Monetary Union, 1992 and [email protected], Salomon Brothers, November. Chubb Corporation, Relative Value Survey of 20 Year General Obligation Bond Trading Values, Warren, New Jersey. CIBC Wood Gundy (1999), ACorporate Spread Roller-Coaster Continues in [email protected], Canadian Financing Quarterly, February 15, Toronto. Economist, The (1999), A Survey of Global Finance: Time for a Redesign?, January 30th-February 5th, London. European Commission (1998a), Communication of the Commission, Financial Services: Building a Framework for Action, Brussels. European Commission (1998b), Proposal for Council Directive to ensure a minimum of effective taxation of savings income in the form of interest payments within the Community, COM (1998) 295/...-EN. European Central Bank (1999), Possible Effects of EMU on the EU Banking Systems in the Medium to Long Term, February, Frankfurt. Feldstein, S.G. and F.J. Fabozzi (1997), AMunicipal [email protected], In: F.J. Fabozzi (ed.) (1997), The Handbook of Fixed Income Securities, Fifth Edition, Irwin Professional Publishing, Chicago, pp. 161-189. Financial Times (1999a), After the Banks, Gillian Tett, Economics Notebook, February 1, 1, p. 16. 29

Financial Times (1999b), Brazil=s Unenviable Choices, Richard Lapped and Geoff Dyer, February 5, p. 16. Financial Times (1999c), Small Countries Pay for Lack of Liquidity, Arkady Ostrovsky and Eric Frey, June 17, p. 27. Flandreau, M., J. Le Cacheux and F. Zumer (1998),AStability without a Pact? Lessons from the European Gold Standard, [email protected], In: D. Begg, J. von Hagen, C. Wyplosz and K.F. Zimmermann (eds.), EMU: Prospects and Challenges for the Euro, Blackwell Publishers, Oxford, pp. 117-162. Goldstein, M. and G. Woglom (1991), AMarket-Based Financial discipline in Monetary Unions: Evidence from the US Municipal Bond [email protected], IMF Working Paper, No. 89, Washington. Goodhart, C.A.E. (1997), AOne Government, One [email protected], Prospect, March, pp. 10-11. Goodhart, C.A.E. (1998), AThe Two Concepts of Money: Implications for the Analysis of Optimal Currency [email protected], European Journal of Political Economy, Vol. 14, No. 3, pp. 407-432. Gravelle, T. (1999), Liquidity of the Government of Canada Securities Market: Stylised Facts and Some Market Microstructure Comparisons to the United States Treasury Market, Financial Markets Department, Bank of Canada, Paper presented for BIS Committee on the Global System Study Group on Market Liquidity. Hagen, J. von (1991), AA Note on the Empirical Effectiveness of Formal Fiscal Restraints, Journal of Public Economics, Vol. 44, No. 2, pp. 199-210. HSBC Markets (1999), AEuroland Sovereigns - A Guide to Credit [email protected], Fixed Income Credit Research, 22 February, London. International Monetary Fund (1998a), Code on Good Practices on Fiscal Transparency Declaration on Principles, 16 April, Washington. International Monetary Fund (1998b), Draft Manual on Fiscal Transparency, Fiscal Affairs Department, 19 October, Washington. International Securities Markets Association (1998), AThe May 1998 EU Proposal on the Taxation of Interest (COM 98 295)@, ISMA Position Paper, September 4, Zurich. Keesing's Record of World Events, London: Cartermill. Lannoo, K. and D. Gros (1998), Capital Markets and EMU, Report of a CEPS Working Party, Centre for European Policy Studies, Brussels. Lemmen, J.J.G. (1998), AIs there Life after the Stability [email protected], FT Single Currency in Practice, December, pp. 10-13. Lemmen, J.J.G. and C.A.E. Goodhart (1999), ACredit Risks and European Government Bond Markets: A Panel-Data Econometric [email protected], Eastern Economic Journal, Vol. 25, No. 1, Winter, pp. 77-107. Lucas, D.J. (1995), ADefault Correlations and Credit [email protected], Journal of Fixed Income, March, pp. 76-78. 30

McCauley, R. N. (1999), "The Euro and the Liquidity of European Fixed Income Markets", Mimeo, BIS, Asian Office. Maastricht Treaty (1992), Treaty on European Union, Council of the European Communities, Commission of the European Communities, Office for Official Publications of the European Communities, Luxemburg, February 7. Mattina, T. and F. Delorne (1997), AThe Impact of Fiscal Policy on the Risk Premium of Government Long-term Debt: Some Canadian @, Department of Finance, Economic and Fiscal Policy Research Working Paper, No. 97-01. McKinnon, R.I. (1997), AMonetary Regimes, Government Borrowing Constraints, and MarketPreserving Federalism: Implications for [email protected], Mimeo, October, Stanford University, Stanford. Millar, J. (1997), AThe Effects of Budget Rules on Fiscal Performance and Macroeconomic [email protected], Bank of Canada Working Paper, No. 15, June, Ontario. Missale, A. (1998), Public Debt Management, Book Manuscript, Forthcoming in Oxford University Press. Obsfeld, M. and K. Rogoff (1996), Foundations of International Macroeconomics, The MIT Press, Cambridge. Massachussets. OECD (1997), Managing Across Levels of Government , PUMA, Paris. Poterba, J.M. (1996), ADo Budget Rules [email protected], NBER Working Paper, No.5550, Cambridge, Massachussets. Poterba, J.M. and K.S. Rueben (1997), AState Fiscal Institutions and the US Municipal Bond [email protected], NBER Working Paper, No. 6237, Cambridge, Massachussets. Queensland Treasury Corporation (1994), Standard and Poor=s Ratings Report on Queensland, 25 January. Siermann, C.L.J. (1996), Politics, Institutions and the Economic Performance of Nations, Ph.D. thesis, Rijksuniversiteit Groningen, Groningen. Stability and Growth Pact (1996), Proposal for a Council Regulation on the Strengthening of the Surveillance and Coordination of Budgetary Positions; Proposal for a Council Regulation on Speeding up and Clarifying the Implementation of the Excessive Deficit and Clarifying the Implementation of the Excessive Deficit Procedure, Commission of the European Communities, COM96 496 Final, Brussels. Standard & Poor's (1998), AGlobal Financial System Stress: Leading Indicators Signal Pressure in 16 [email protected], Features, Credit Week, September 16, pp. 9-15. Stiglitz, J. and A. Weiss (1981), ACredit Rationing in Markets with Imperfect [email protected], American Economic Review, Vol. 71, No. 3, pp. 393-410. Sutter, M. (1999), Voting and Voting Power in the Stability Pact, Institute of Public Finance, Mimeo, University of Innsbruck, Innsbruck, February. 31

World Bank (1995), Brazil State Debt: Crisis and Reform, November 14, Report No. 14842-BR. World Bank (1996a), Argentina Provincial Finances Study, Selected Issues in Fiscal Federalism, Volume I: The Main Report, July 12, Report No. 15487-AR. World Bank (1996b), Argentina Provincial Finances Study, Selected Issues in Fiscal Federalism, Volume II: Technical Annexes, July 12, Report No. 15487-AR.

32